Skip to content

Roadmap

This page describes what was built in each release and what is planned for future releases. The [Unreleased] section of CHANGELOG.md contains the most up-to-date view of in-progress work.

Released

v0.1 — Core

Foundation: configuration, transport, and routing.

  • RabbitConfig — composable configuration dataclasses, frozen and slot-based
  • AsyncBroker and SyncBroker — async (aio-pika) and sync (pika) broker implementations
  • @subscriber and @publisher decorators — declarative routing and handler registration
  • Basic ack policies — AUTO (ack on success, nack on exception)
  • Topology declaration — exchange and queue declared from RabbitExchange / RabbitQueue at startup

v0.2 — Reliability

Production-grade message safety.

  • RetryConfig — configurable max retries and per-attempt delay sequence
  • Dead-letter queue support — automatic DLQ topology declaration and routing
  • Publisher confirms — all retry and DLQ publishes confirmed before original ack
  • AckPolicy.MANUAL — full settlement control in handler code
  • AckPolicy.NACK_ON_ERROR — explicit nack with requeue on handler exception

v0.3 — Testing

In-process testing without a running RabbitMQ.

  • TestBroker — in-memory broker that routes through the real handler pipeline
  • In-memory routing — publish and subscribe wired together without transport layer
  • pytest fixtures — test_broker fixture for use in test functions and conftest

v0.4 — Production Middleware and Observability

Middleware layer and Kubernetes lifecycle.

  • Middleware base class — composable pre/post handler hooks
  • DeduplicationMiddleware — exactly-once delivery via Redis-backed nonce store
  • RateLimitMiddleware — per-consumer rate limiting with token bucket
  • CompressionMiddleware — gzip/zstd/lz4 transparent compression and decompression
  • SigningMiddleware — HMAC-SHA256 message signing and verification
  • MetricsMiddleware — Prometheus counters and histograms for handler latency and outcomes
  • TimeoutMiddleware — per-handler deadline with configurable action on expiry
  • Health checks — broker_health_check() and broker_health_check_async() for liveness and readiness
  • Kubernetes lifecycle — graceful shutdown on SIGTERM, drain period, readiness gating

v0.5 — RabbitMQ Operations

CLI tooling, API ergonomics, and operational observability.

  • CLI — rabbitkit dlq inspect/replay, rabbitkit health liveness/readiness, rabbitkit topology list/validate/diff/apply, rabbitkit routes list/describe
  • topology validate / topology diff — compare declared topology against live RabbitMQ via management API
  • topology apply — declare all registered queues and exchanges via AMQP (with --dry-run)
  • rabbitkit.aio — clean alias for rabbitkit.async_; both paths are identical and supported
  • rabbitkit.experimental — groups RPC, locking, signing, streams, result backends, and dashboard under one import namespace
  • Simple publish API — broker.publish(routing_key=..., body=..., headers=...) without requiring MessageEnvelope
  • DeduplicationConfig(mark_policy=) — "on_success" (default, safer for retries) or "on_start" (prevents concurrent duplicates)
  • Benchmarks — real RabbitMQ (testcontainers) for throughput, latency, failure, resource, sync, lifecycle; result persistence with regression comparison
  • Production examples — publish_message, pydantic_validation, publisher_confirms, kubernetes_worker (with K8s manifest, health server, SIGTERM drain)
  • Dashboard — web UI for live queue and consumer state inspection (experimental)
  • Management client — programmatic access to the RabbitMQ Management HTTP API
  • Stream queue support — StreamOffset, StreamConsumerConfig, StreamOffsetType for RabbitMQ 3.9+ streams

Planned

v1.0 — Stable API

API freeze, full documentation, full type coverage, and migration support.

  • Stable API freeze — all symbols listed in the stability policy are frozen; semver guarantees apply from this release
  • Full documentation — complete docs for all stable APIs, guides for common patterns, and architecture reference
  • Full type coverage — mypy --strict clean across the entire public API surface
  • Migration guide — documented upgrade path from 0.x to 1.0 for any breaking changes introduced during the 0.x stabilization period

Under consideration (unscheduled)

  • SASL EXTERNAL authentication (x509 certificate auth) — for deployments that authenticate clients by their TLS client certificate instead of username/password. SecurityConfig(mechanism=...) currently fails fast for anything other than PLAIN (deliberately — a silent fallback would be worse). Both pika (pika.credentials.ExternalCredentials) and aio-pika support it, so this is plumbing plus tests, prioritized on demand — open an issue if your deployment needs it.

Not on the Roadmap

The following are out of scope for RabbitKit's intended design:

  • Multi-broker support (Kafka, NATS, Redis Streams) — use FastStream for multi-broker workloads
  • A general-purpose task scheduler — use Celery Beat for cron-style task scheduling
  • A result backend as a primary feature — Celery's result backend is more mature; RabbitKit's result backend support remains experimental and scoped to RPC use cases